CHAPTER 9

INFANT AND CHILD MORTALITY

Tauseef Ahmed, Mansoor-ul-Hassan Bhatti and George Bicego

Since independence in 1947, Pakistan has experienced steady but modest declines in rates of mortality at all ages. Yet early childhood mortality remains high. The most recent source of child survival data before the 1990-91 PDHS was the multi-round Pakistan Demographic Survey (PDS), which estimated infant mortality at 110 per thousand live births for the 1984-88 period. The PDHS provides an opportunity to examine infant and child mortality in more detail. Such information is important at this time because of recent efforts by the health sector in Pakistan to address the health problems of infants and young children. Major areas of child survival intervention include: (1) control of diarrhoeal disease, (2) management of acute respiratory infection, (3) the Expanded Programme on Immunisation (EPI), and (4) nutrition. Careful examination of changes in the level of infant and child mortality is essential in assessing whether and to what extent such efforts have been effective.

This chapter focuses on levels, trends, determinants and differentials in the mortality of infants and children under the age of five. An analysis of high-risk fertility behaviour is also presented.

9.1 Definitions of Infant and Child Mortality

All female respondents were asked to provide a complete birth history, including the sex, birth date, survival status, and current age or age at death for each live birth. These data were used to calculate the following direct estimates of infant and child mortality, 1 for three six-year periods (0-5, 6-11, and 12-17) preceding the survey:

Neonatal mortality: the probability of dying in the first month of life; Postneonatal mortality: the difference between infant and neonatal mortality; Infant mortality (lq,): the probability of dying before the first birthday; Child mortality (4ql): the probability of dying between the first and fifth birthday; Under-five mortali ty (sq0): the probability nf dying before the fifth birthday.

i A detailed description of the method for calculating the probabilities presented here is given in Rutstein (1984). The mortality estimates are not rates, but are true probabilities, calculated according to the conventional life table approach. For any calendar period, deaths and exposure in that period are first tabulated for the age intervals 0, 1-2, 3- 5, 6-11,12-23,24-35, 36-47, and 48-59 months. Then age interval specific probabilities of survival are calculated. Finally, probabilities of mortality for larger age segments are produced by multiplying the relevant age interval survival probabilities together and subtracting the product from one:

|ffiZ÷n

,,q~, = 1-1-I (1-q~) lmZ

111

9.2 Assessment of Data Quality

Rates of infant and child mortality are subject to both sampling and nonsampling errors. This section describes the results of some basic checks for various nonsampling errors; namely, underreporting of early childhood deaths (which would result in underestimates of mortality) and misreporting age at death (which may distort the age pattern of under-five mortality).

Underreporting of deaths is generally more common for children who died shortly aRer birth than those who died later. If early neonatal deaths are selectively underreported, then an abnormally low ratio of deaths under seven days to all neonatal deaths and an abnor- mally low ratio of neonatal to infant moaality would be observed. If such underreporting is related to the child's sex orto the length of time preceding the survey, then the ratios would be affected in proportion to the extent of the underreporting. The ratios are given in Table 9.1.

The proportion of neonatal deaths reported to have occurred during the first week of life (0-6 days) varies considerably for males, but little for females. The large jump in the ratio for males between 12-17 years before the survey and 6-11 years before the survey from 0.50 to 0.69 is unusual and suggests that male deaths during the early neonatal period may have been omitted in the earliest period. The ratio of neonatal mortality to infant mortality is comparatively stable for beth males and fe- males although further evidence in support of the notion of male omission for the earliest period is found in a slight increase in the male ratio between the two earliest periods, when the female ratio was declining. Since female neona- tal mortality is expected to be lower due to greater biological vigour at birth, it is unusual to find a smaller male neonatal/infant ratio for the earliest period.

Table 9.1 Indices of early infant deaths

Indices of early infant deaths by sex, for three six-year periods, pakistan 1990-91

Time period of death (years preceding survey)

Index 0-5 6-11 12-17

Deaths in first 6 days/ all neonatal deaths

Male 0.64 0.69 0.50 Female 0.62 0.62 0.59

Neonatal deaths/ all infant deaths

Male 0.59 0.61 0.59 Female 0.55 0.55 0.60

Table 9.2 Sex differentials in infant and neonatal mortality

Infant and neonatal mortality rates by sex, and sex ratios, for three six-year periods preceding the survey, Pakistan 1990-91

Time period of death (years preceding survey)

Mortality 0-5 6-11 12-17

Infant mortality Male 100.9 98.5 101.6 Female 79.7 93.9 113. l Sex ratio 1.27 1.04 0.90

Neonatal mortality Male 59.0 60.2 59.4 Female 43.5 53.1 67.7 Sex ratio 1.36 1.13 0.88

Sex differentials in infant and neonatal mortality over time are presented in Table 9.2. By analysing trends in these differentials, it is possible to examine misreporting of mortality. Normally, there is an excess of male mortality during infancy, especially during the neonatal period. The decreasing ratio of male to female mortality going back in time is striking, and suggests that some early death.s of male children 12-17 years preceding the survey were not reperted?

2 On the other hand, female mortality may have been decreasing at a faster pace than male mortality, which could partially account for the unusual pattern seen here.

112

This would lead to a bias in the evaluation of trends, since mortality rates 12-17 years preceding the survey win be underestimated relative to roortality in more recent time periods.

A problem common to roost retrospective surveys is heaping of age at death on "convenient" digits, e.g., 6, 12, and 18 months. This phenomenon introduces biases in rate calculation, if the net result is to shift deaths from one age segment to another. Thus, heaping at 12 months causes concern because a certain fraction of these deaths, though reported to occur after infancy (i.e., at ages 12-23 months), may have actually occurred during infancy (i.e., at ages 0-11 months). The infant mortality rate (lqo), in this case, is biased downwards and child mortality Gq~) upwards.

Table 9.3 presents the distribution of deaths reported at ages 5 to 23 months by reported age at death for three six-year periods preceding the survey. Distinct "heaps" of deaths are evident at 6, 12, and 18 months of age, with corresponding deficits in the adjacent months, although heaping at 12 months occurs only during the roost recent period. Digit preference appears not to be serious enough, however, to substantially alter the rates calculated here. For instance, even if as many as half of the deaths reported at "12 months" were reassigned to the infant age segment, infant mortality would be increased and child mortality would be decreased by less than one percent for the period 0-5 years preceding the survey. More troublesome are the large number of deaths re- ported at "1 year," which results not from digit pref- erence, but from the failure of interviewers to correctly elicit and record the age at death in units of months as required by PDHS survey guidelines for deaths under two years o f age? The rates reported here are not adjusted for this defect in the data----which, in effect, assumes that all "1 year" deaths occurred in the 12-23 month segment. In the unlikely event that half of these deaths actually occurred during the infant age segment, the infant mor- tality rate would be underestimated by less than five percent for each of the three six-year periods. While in- fant mortality may be slightly underestimated and child mortality overestimated, when age at death is reported as "1 year" (with no adjustment), trend evaluation will not be substantially affected, since there is no marked trend in such reporting.

Unreported age at death is another potential prob- lem in data of this type; however, respondents failed to provide age at death in only 5 (0.2 percent) of the 3,016

Table 9.3 Reporting of age at death

Distribution of reported deaths at age 5-23 months by age at death, for three six-year periods preeedin_g the sta'vey, Pakistan 1990-91

Time period of death (years preceding survey)

Age at death (months) 0-5 6-11 12-17

5 16 28 15 6 39 38 22 7 13 21 20 8 18 17 10 9 21 9 24 10 18 21 12 11 9 9 4 12 20 7 5 13 2 0 0 14 1 3 1 15 10 4 3 16 5 0 3 17 2 1 0 18 16 17 10 19 0 0 0 20 0 0 0 21 l 0 l 22 1 0 0 23 0 0 0 1 year I 69 61 55

1Age at death reported to be 1 year was supposed to be recorded in months.

3 It can be argued that these deaths are more likely to be drawn from the 12-23 month age segment than deaths reported at "12 months," which presumably are drawn from beth older and younger ages (reflected in the deficit at 11 months). This is based on the notion that a report of "1 year" would commonly translate to having reached the first birthday (i.e., one completed year 0¢ 12-23 months). Any adj us tment procedure that would involve teas sign ing a fraction of these deaths from the post-infant to infant age segment is, thus, undertaken with a great deal of uncertainty.

113

deaths reported during the 18 years preceding the survey. In these cases, age at death was imputed using a hot-deck procedure. 4

In sum, internal consistency checks indicate that the mortality data from the PDHS are of reasonable quality, keeping in mind the sociocultural constraints prevalent in the society. However, with increasing time since the survey, there appears to be an increasing omission of neonatal deaths, especially deaths to boys. The effect of this omission will be to underestimate rates of infant mortality for less recent periods and thus reduce estimates ofmortalitydecline. Also, imprecisereportsofages atdeath, especially reports o f " l year," may introduce a small downward bias in the estimation of infant mortality and an upward bias in the estimation of child mortality. However, the evaluation of trends in the PDHS would not be affected.

9.3 Levels and Trends in Infant and Child Mortality

Infant and child mortality rates for three six-year periods preceding the PDHS are shown in Table 9.4. The estimated infant mortality rate for the most recent period (0-5 years preceding the survey) is 91 per 1000 live births, with 57 percent of infant deaths occurring in their first four weeks of life. 5 Under-five mortality (sqo) has fallen 18 percent, from 143 to 117 per thousand in the period from 12-17 years to 0-5 years before the survey. Much of the estimated decline in mortality occurred between the periods 12-17 years and 6-11 years preceding the survey, with a slower rate of decline between the two most recent periods. Dis- proportionate gains were made in survival among children age 12 to 59 months (28 percent), while neonatal (19 percen0 and postneonatal (11 percent) survival have shown more modest improvements. As mentioned previously, the apparently unremarkable change in neonatal mortality may be an artifact of underreporting of neonatal deaths in less recent periods. Still, over half of infant mortality (neonatal plus postneonatal mortality) and 45 percent of under-five mortality occurs during the first month of life (see Figure 9.1). Thus, there exists considerable scope for improving child survival during infancy in Pakistan through programmes designed to improve maternal health and pregnancy outcome.

Table 9.4 Infant and child mortality

Infant and child mortality rates for three six-year periods preceding the survey, Pakistan 1990-91

Neonatal Postneonatal Infant Child Under-five Years preceding mortality mortality I mortality mortality mortality survey (NN) (PNN) (lqo) (4qt) (sqo)

0-5 51.4 39.1 90.5 29.5 117.4 6-11 56.7 40.0 96.7 29.8 123.6 12-17 63.4 43.7 107.1 40.7 1,~3.5

Note: The month of interview is excluded from the analysis. 1Computed as the difference between infant and neonatal mortality

4 The procedure assigns an age at death equal to that of the last death of the same birth order in the data file. The estimated infant mortality rate based on births and deaths for a period of 12 complete months preceding the

1984-85 PCPS was 106.

114

Figure 9.1 Distribution of Deaths

Under Age Five by Age at Death

Neona ta l 45%

Id (1-4) 23%

Note, Based on death~ in the s i x years preceding the survey,

PDHS 1990 -9 t

9.4 Differentials in Infant and Child Mortality

Table 9.5 presents infant and child mortality rates by urban-rural residence, province of residence, level of mother's education, and use of basic maternal health services. Mortality rates are calculated for a ten-year period so that the rates for each population subgroup are based on an adequate number of events. Under-rive mortality (sq0) is 29 percent lower in urban Pakistan (94 per thousand) than in the rural setting (132 per thousand). The urban-rural differential exists at all ages (see Figure 9.2), which suggests that both social factors and access to health services are important in the greater risk of death among rural children. There is little or no difference in mortality risk between major cities and other urban areas.

Among regions, under-five mortality is lowest in NWFP (98 per thousand) and highest in Punjab (133 per thousand). The higher under-five risk in Punjab is largely due to higher mortality during infancy (104 per thousand). Infant mortality was estimated to be lowest in Balochistan, but this should be interpreted with caution since sampling errors are relatively high in Balochistan, as well as in NWFP. Moreover, the unusually low rate of infant mortality relative to child mortality in Balochistan as well as a deficit of births in the year preceding the survey, suggests that there was selective underreperting of infant deaths in that province,

Child survival chances in Pakistan are closely related to a mother's level of education. Children of mothers with no education experience over two and a half times (159 percent) the level of under-five mortality as children of women educated to the secondary level or higher. Indeed, each incremental change in education is associated with significant gains in survival. In the posmeonatal period, a strong education effect is observed only for the secondary and higher level of education.

115

Table 9.5 /nfant and child mortality by background characteristics

Infant and child mortality rates for the tan-year period preceding the sut~ey, by background charactexlsfxcs of ~ e mother, Pakistan 1990-91

Neonatal Postnannatal Infant Child Under-five Characteristic mortality mortality I mortality mortality mortality

of mother (NN) (PNN) Qqo) (4ql) (~o)

Residence Total urban 40.8 33.8 74.6 20.6 93.6

Major city 39.7 34.1 73.8 19.5 91.9 Other urban 42.1 33.5 75.6 22.0 95.9

Rural 58.6 43.7 102.2 33.0 131.9

Province Punjab 58.4 45.7 104.1 32.0 132.8 Sindh 44.4 36.1 80.5 27.3 105.6 NWFP 48.2 31.3 79.6 19.7 97.7 Balochistan 46.1 26.3 72.4 31.0 101.l

Education level aRended No education 56.1 42.5 98.6 33.1 128.4 Primary 49.8 40.6 90.4 18.3 107.0 Middle 43.5 36.7 80.2 7.9 87.4 Secondary + 26.9 18.9 45.8 3.9 49.5

Medical maternity care z No antenatal or 57.1 47.2 104.3 31.6 132.6 delivery care

Either antenatal or 32.5 32.4 64.9 40.9 103.1 delivery care

Both antenatal and 46.7 22,8 69,5 10,5 79.3 delivery care

Total 53.3 40.7 94.0 29.2 120.4

Note: The month of interview is excluded from analysis. IComputed as the difference between infant and neonatal mortality ~Rates are for the five-year period preceding the survey. Medical care is that given by • doctor, nurse or lxained midwife, or rece'tved in a hospital, clirftc, health centre or health unit.

116

70

60

SO

40

30

20

10

0

Figure 9.2 Infant and Child Mortality

by Place of Residence

Deaths pe r 1,000

Total Major City Other Urban Rural

Note: Based on deaths in the 10 years preceding the survey.

PDHS 1990-91

Use of basic maternal health services is associated with child survival chances. Under-five mortality is 67 percent higher ( 133 per thousand) among children born to women who did not receive antenatal services and did not receive delivery care from a trained health professional compared to children whose mothers received both services. The under-five differentials, however, exhibit some unexpected differences in mortality risk across service use categories. For instance, use of both services is associated with higher neonatal risk than use of one or the other. Perhaps, women who receive antenatal care do not also seek delivery care unless they experience a problem with the pregnancy. Child mortality (4qt) is lower among children whose mothers received no services than among those whose mothers received one or the other service. This aberrant pattern may be the result of small numbers, which produce large sampling errors. Nevertheless, it is unusual that the postneonatal period, and not the neonatal period, is the age segment most sensitive to the effects of maternal health service use.

Table 9.6 presents differentials in infant and child mortality by various characteristics of the mother and child. The expected biological effects of sex on age-specific mortality are observed. Neonatal mortality is 30 percent higher among males than females; however, mortality during the posmeonatal period is little affected by the sex of the child. Child mortality (,~ql) is 66 percent higher among females than males, which suggests that there may be some gender-related differences in child rearing practices that favour boys over girls.

117

Table 9.6 Infant and chiM mortality by demographic characteristics

Infant and child mortality rates for the ten-year period preceding the survey, by selected demographic and biological cliaracteristics, Pakistan 1990-91

Demographic/ Neonatal Post.neonatal Infant Child Under-five biological mortality mortality l mortality mortality mortality characteristic (NN) (PNN) (lqo) (4ql) (sqo)

Sex of child Male 60.1 42.0 102.1 22.0 121.9 Female 46.1 39.3 85.5 36.5 118.9

Mother's age at birth <20 70.1 51.2 121.3 26.7 144.8 20-29 50.9 40.0 90.8 28.5 116.7 30-39 48.5 35.4 83.9 31.8 113.0 40-49 56.1 50.4 106.5 27.1 130.7

Birth order 1 60.9 37.7 98.5 12.9 110.2 2-3 45.0 38.6 83.6 32.1 113.0 4-6 51.9 43.0 94.8 33.6 125.2 7+ 63.4 43.5 106.9 32.3 135.7

Previous birth Interval < 2 years 74.1 58.8 132.9 42.9 170.1 2-3 years 38.9 26.3 65.2 25.5 89.0 4 years or more 13.5 16.3 29.7 15.0 44.3

Birth size 2 Very small 90.6 40.7 131.3 30.5 157.8 Small 41.9 58.0 99.9 22.9 120.5 Average or larger 40.3 31.5 71.8 27.6 97.4

Note: The month of interview is excluded from analysis. tComputed as the difference between infant and neonatal mortality 2Rates are for the five-year period before the survey.

Differences in under-five mortality (sq0) by matemal age at birth follow the expected pattern: mortality is highest for children of young mothers; it falls for births to mothers age 20-39; then it rises again for births to women 40 and older (see Figure 9.3). This pattern, however, is seen exclusively during the first year of life. This same U-shaped pattern is also seen in the relationship between birth order and mortality, but only during the neonatal period. After the first month, first order births are at lowest risk, and the high birth order effect is not observed.

The pace of childbearing has a powerful effect on the survival chances o f Pakistani children. Under- five mortality (.41o) is four t imes higher among children born after an interval of less than two years than among children born after an interval o f four years or more. The birth interval effect is marked for mortality in each age group, although the strength of the association diminishes with increasing age of the child. Birth intervallength strongly affects survival chances throughout the first five years of life. This may indicate that

118

80

70

60

50

40

30

20

10

0

Figure 9.3 Infant Mortality by Mother's Age

at First Birth

Deaths per 1,000 B i r ths

,20 20-29 30-39 40-49

Mother's Age

1 N e o n a t a l m Postneonatal

PDHS 1990-91

the relationship in Pakistan is not simply related to maternal depletion and pregnancy outcome (which would be expected to specifically influence early infant mortality), but may also be associated with constraints on breastfeeding and other nutritional inputs, child care, and the use of health services.

The size of a child at birth is closely linked to neonatal survival. Children perceived by their mothers as very small were 2.2 times more likely to die in the first four weeks than those perceived as small, average or larger. After the neonatal period, the relationship is less pronounced, although average~larger children still appear to maintain some continued advantage in survival in the postneonatal period.

9.5 Causes of Death in Early Childhood

In the PDHS, an effort was made to obtain general information on the causes of death of children born during the five-year period preceding the survey. Mothers of deceased children were asked whether a doctor or health care worker had provided an explanation or cause for the child's death. If the mother responded "yes," then she was asked to give the reported causes of death. If there was no explanation from a doctor or health worker, then the mother was asked to provide her perception of the causes of death.

Table 9.7 gives the percentage of deaths in the five years preceding the survey by reported causes of death, the source of information on the cause, and the age group of the child at death. The figures should be viewed with caution, especially those for children age 12-59 months, since they are based on small numbers. Furthermore, nearly half of all reported deaths could not be assigned to any specific category, making interpretation difficult. This is particularly true for reporting of causes during the neonatal period, when many deaths are attributed to a lack of viability at birth and, as such, are difficult to specify. This is further reflected in the small fraction of deaths for which a doctor or health worker's explanation of the cause of death could be elicited (26 percent during the neonatal period compared to 42 percent during the posmeonatal period and 55 percent for children age 12 to 59 months).

119

Table 9.7 Causes of death in early childhood

Among nonsurviviug children born dusting the five years preceding the survey, the percentage reported to have died from selected diseases or other causes, by age at death and person reporting the cause of death, Pakistan 1990-91

Age at death

Less than 1 month 1-11 months 12-59 months Total

Cause reported by Cause reported by Cause reported by Cause reported by

Disease or cause Doctor I Mother z Total Doctor I Mother 2 Total Doctor t Mothe~ Total Doctor 1 Mother 2 Total

Diarrhoea 0.6 7.2 5.5 40.2 19.5 28.3 (30.0) (22.4) 26.6 24.1 12.8 16.9 Vonfitiug 2.2 5.6 4.7 7.2 13.1 10.6 (11.7) (19.1) 15.0 6.5 9.5 8.4 Pneumonia 7.4 2.5 3.7 16.9 0.7 7.5 (5.7) (--) 3.1 11.0 1.6 5.0 Cough 1.6 4.4 3.7 8.8 10.9 10.0 (5.4) (17.5) 10.8 5.5 7.9 7.0 Fever 19.9 14.5 15.9 33.3 40.8 37.6 (30.4) (49.6) 39.0 28.0 26.8 27,2 Convulsions 5.2 8.6 7.7 1.8 8.5 5.7 (8.1) (8.4) 8.2 4.4 8.5 7.0 Measles 4.0 1.3 2.0 5.2 1.7 3.2 (14.3) (7.0) 11.1 6.9 2.0 3.8 Other 63.8 68.6 67.4 16.2 43.6 32.0 (27.7) (33.0) 30.1 35.4 56.7 49.0

Number 69 200 268 84 114 198 45 37 82 198 350 548

Note: Percentages add to more than 100.0 because of the recording of multiple causes of death. Figmes in parentheses are based on 25 to 49 unweighted cases. -- Less than 0.05 percent 1Reperted to mother by a doctor or a health waker 2Mother's own ~ssessrnent (for mothers who were not informed about the cause by a doctor or a health worker)

Despite these problems, some conclusions can tentatively be drawn from the figures on causes of death in childhood. First, over one-quarter (27 percent) of the under-five deaths were associated with fever, and about 17 percent were associated with diarrhoea (as reported either by the mother or a health worker). Cough, vomiting, convulsions, pneumonia, and measles were each reported for 4 to 8 percent of deaths, although in the case of measles, it is unlikely that the few deaths reported during the neonatal period were actually associated with measles. Deaths associated with reported convulsions may represent cases of tetanus during the neonatal period, and aRer the neonatal period often would suggest various conditions involving high fever or hypoxia such as meningitis, pneumonia, and other acute infections. During the posmeonatal period, there is a two-fold difference between mothers' reports and health workers' reports in the percentage of deaths associated with diarrhoea. This difference suggests that children with the most severe cases of diarrhoea tend to be brought to doctors for treatment.

Lastly, it was mentioned above that the percentage of child deaths for which a doctor's or health worker's explanation was provided rises markedly with the age group of the deceased child. One substantive expla-nation for this pattern reflects greater use of curative health services among older children, which would increase the probability of the child having seen a health worker around the time of death, in addition, illnesses among older children may be less serious but of longer duration, thereby increasing the probability of having had some comact with health personnel.

120

In sum, acute respiratory and gastrointestinal illnesses are important morbid conditions preceding death al~ter the neonatal period in Pakistan. Causes of neonatal mortality are more difficult to ascertain with these very limited data, although it appears likely that neonatal tetanus and other acute infectious diseases play at least some role.

9.6 High-risk Fertility Behaviour

Previous demographic research has shown that a strong relationship exists between a mother's pattern of fertility and her children's survival chances. Typically, infants and young children have a higher risk of dying if they are born to very young mothers or older mothers, if they are born after a short birth interval, or if their mothers have already had many children (see Table 9.6). In the following analysis, mothers are classified as too young if they are less than 18 years old at the time of the birth, and too old if they are 35 or older at the time of the birth. A short birth interval is defined as less than 24 months, and a high order birth as one occurring after five or more previous births (i.e., birtb order six or higher). Births may also be cross- classified by combinations of these characteristics. Thus, a birth may have from zero to three high-risk characteristics.

Column one of Table 9.8 shows the percentage of births in various risk categories. More than half (55 percent) of all births in Pakistan fall into at least one risk category, with a third (19 percent) of these having multiple risk characteristics. Risk ratios are presented in column two; a risk ratio is the ratio of the proportion of children in a category who have died, to the proportion who have died in the reference category of births without any risk factors. This analysis indicates several things. First, older maternal age at birth (age >35) and high parity (BO >5) are essentially unrelated to mortality risk when not combined with a short birth interval (BI <24). Indeed, the two main factors leading to heightened risk are young age at birth (< 18 years) and a short birth interval (<24 months). The only multiple risk categories that exhibit higher risk are those that include a short birth interval. Since one-third of recent births in Pakistan occurred less than 24 months after a prior birth, this fmding underscores the need to reduce, through greater use of contraception, the number of closely spaced births.

Column three in Table 9.8 shows the distribution of currently married women by the risk category into which a current birth would fall. A comparison of this percent distribution with the distribution of actual births in the last six years indicates that the percentage of births falling into two of the high-risk categories would increase substantially without fertility control. However, both of these categories (births to older women and births to older women of higher parity) have risk ratios of less than one and, therefore, they do not conform to the expected pattern.

121

Table 9.8 High-risk fertility behavioor

Percent distribution of chil&en loom in the six years preceding the survey who are at elevated risk of mortality, and the percent distribution of currently married women at risk of conceiving a child with an elevated risk of mortality, according to category of increased risk, Pakistan 1990-91

Births in the 6 years preceding the survey Percentage of

currendy Risk Percent of Risk married category births ratio women a

Not in any risk category 45.1 1.00 33.3 b

Single r isk category Age<18: Mother 's age at birth <18 3.8 Age>34: Mother's age at birth 35+ 2.7 BI<24: Birth interval <24 months 19,9 BO>5: Birth order 6 or higher 9.4

1.69 1.0 0,93 8.8 1,78 17.2 0.93 7.4

Subtotal 35.9 1.48 34.4

Multiple risk category Age<18 & BI<24 c 1.0 1.64 0.3 Age>34 & BI<24 0.5 1.17 0.8 Age>34 & BO>5 7.0 0.81 20.8 Age>34 & BI<24 & BO>5 3.1 2.01 4.7 BI<24 & BO>5 7.5 2.25 5.7

Subtotal 19.0 2.00 32.0

In any r isk category 54.9 1.53 66.7

Total 100,0 NA 100.0 Number 8241 NA 6364

Note: Risk ratio is the ratio of the proportion dead of births in a specific risk category m the proportion dead of births not in any risk category. NA = Not applicable aWomeaa were assigned to risk categories according to the status they would have at the birth of a child, if the child were conceived at the time of the survey: age less than 17 years and 3 months, age older than 34 years and 2 months, latest birth less than 15 months ago, and latest ~ of order 5 or higher. h i ' h i d e s sterilised women C[ncindes the combined categories age <18 and birth order >5,

122

REFERENCES

Federal Bureau of Statistics [Pakistan]. 1990. Pakistan Demographic Survey - 1988. Karachi.

Rulstein, Shea Oscar. 1984. Infant and Child Mortality: Levels, Trends, and Demographic Differentials. Revised edition. WFS Comparative Studies No. 43. Voorburg, Netherlands: International Statistical Institute.

123